An Improved ACO based on Estimation of Distribution

Abstract

In this Thesis a routing algorithm named Ant Routing System (ARS) is presented. ARS is based on a general-purpose met heuristic named “An Improved ACO based on Estimation of Distribution” which is a framework for building ant-inspired algorithms. ARS is applied as the routing algorithm in a point-to-point network Estimation of Distribution. ED-ACO yields significant improvements in performance over ACS. Ant Colony System algorithm is one of the best algorithms of Ant Colony optimization however, the weaknesses of premature convergence, slow speed and low efficiency greatly restrict its application. An improved method to deal with the two main problems involved in ACO, slow speed of convergence and poor ability to search better solution at the end of the search procedure. In order to improve the performance of the algorithms, a new Ant Colony Optimization algorithm based on Estimation of distribution (ED-ACO) is presented. ED-ACO is significantly improving the performance of slow speed & premature convergence. The MATLAB results show that ED-ACO is an effective and efficient way to solve combinatorial optimization problems. ARS based on ED-ACO behaves differently depending on the relative priority of positive feedback, negative feedback and local heuristics, and that it is possible to adjust the parameters to achieve distribution of traffic over several paths when the network is heavily loaded, resulting in a higher throughput and lower loss. In this Thesis, I have implement ED-ACO on MATLAB and show the improvement in the performance such that convergence rate become faster compared to other ACO based algorithm and also increases the error performance.